<  February 2017   >
Su Mo Tu We Th Fr Sa  
          1  2  3  4  
 5  6  7  8  9 10 11  
12 13 14 15 16 17 18  
19 20 21 22 23 _2_4 25  
26 27 28
01:09 blueness joined
01:19 blueness joined
02:16 <andypost> jirutka, ncopa Here's a all php extensions (tested everything that can) https://goo.gl/w2qgEz
02:16 <jirutka> andypost: it requires me to login…
02:17 <andypost> https://goo.gl/w2qgEz
02:17 <andypost> shared
02:17 <jirutka> better
02:32 <andypost> jirutka, so mostly all php7* are compatible with 7.1
02:48 s33se_ joined
03:40 deepbook5broo joined
03:40 deepbook5broo left
04:59 <scadu> anyone knows if Austin Page is here?
05:02 <scadu> py-cassandra-driver build fails on subpackaging. http://tpaste.us/jeDK
05:04 <scadu> arch has been defined in _py section and I wonder how it should be handled nowadays.
05:13 <scadu> still not sure why arch has been defined in _py function. sorry, I just woke up seing that rebuild failed on this :P
05:14 <scadu> seeing even
05:36 skarnet joined
05:55 fabled joined
06:21 czart joined
06:42 minimalism joined
07:07 vakartel joined
07:14 <fabled> i wonder what is llvm state currently and if it's soon time to look at it as main toolchain or not
07:48 t0mmy joined
08:27 stwa joined
08:42 hd joined
09:06 tmh1999 joined
09:13 <ncopa> there was a new llvm release
09:13 <ncopa> andypost: nice list! thank you
09:19 fekepp joined
09:51 royger joined
09:53 vakartel joined
09:54 stwa joined
10:12 <ncopa> boost-python3 is broken
10:12 <ncopa> for some reason it links against python2
10:12 <ncopa> which is weird
10:14 <scadu> oO
10:14 <scadu> and I'm still doing rebuild locally
10:14 <ncopa> seems they have same issue on archlinux
10:14 <ncopa> https://git.archlinux.org/svntogit/packages.git/tree/trunk/PKGBUILD?h=packages/boost#n64
10:46 <tmh1999> fabled : in scripts/bootstrap.sh, after msg "Cross building base system", we should add gcc, g++, gnat for $TARGET_ARCH, right ?
10:50 ncopa joined
10:55 <ncopa> chromium crashes again
10:58 <ncopa> fabled: i think i have a possible apk bug
10:58 <ncopa> to reproduce: apk search --exact -r so:libvncclient.so.0
10:58 <ncopa> it should have shown remmina
10:59 <ncopa> but it doesnt
10:59 <ncopa> $ apk info -R remmina | grep libvncclient
10:59 <ncopa> so:libvncclient.so.0
11:01 <ncopa> $ apk dot --errors | grep libvncclient
11:01 <ncopa> should also shown aqemu
11:20 leitao joined
11:36 <fabled> jirutka, http://sprunge.us/YCRS looks ok?
11:44 <fabled> bummer, ghc was bootstrapped with wrong triplpet
11:45 <ncopa> how to generate a graph of edge main+community repo errors without messing with local repositories:
11:45 <ncopa> docker run --rm -it alpine:edge sh -c "apk update -q --no-progress -U && apk dot --error" | dot -Tsvg > output.svg
11:49 <andypost> looks md5&sh256 checksums are deprecated, new abuild?
11:52 <ncopa> yeah, we dont generate them anymore
11:52 <ncopa> to avoid cluttering the diffs
11:52 <ncopa> if they are in APKBUILD they will be verified
11:53 <ncopa> we just stop generate them
11:53 <_ikke_> What if you update a package?
11:53 BitL0G1c joined
11:53 <_ikke_> would apk checksum still update the existing sums?
11:53 <ncopa> will remove md5 and sha256
11:54 <_ikke_> ah ok
11:54 <ncopa> and only add sha512
11:54 <ncopa> md5 is kinda meaningless nowdays
11:55 <ncopa> what we should do though, is make abuild verify gpg signatures if provided
11:55 <ncopa> i wonder what broke chromium
11:57 <^7heo> rust?
11:57 <* ^7heo> hides
11:59 <ncopa> could be nss
12:00 LouisA joined
12:00 <^7heo> (I know it's not ##physics, but really nobody got my joke?)
12:01 <ncopa> apparently not :)
12:03 <^7heo> chromium is a metal, and rust is another name for corrosion; a reaction that usually degrades (or break metals)
12:04 <^7heo> now the joke is: firefox, another browser, has recently got a dependency on rust (lang) while chromium (the browser) has the same name as the element, which is rust resistant (haha)
12:06 <^7heo> so basically it goes down to parallel between chemistry/physics and software ecosystems
12:07 <^7heo> and the politics in browsers
12:07 <^7heo> and derivated languages
12:10 <^7heo> and as a correlated joke: go rust!
12:25 <_ikke_> ^7heo: I did get the joke but just did not respond
12:25 <^7heo> _ikke_: fair enough :P
12:26 farosas joined
12:36 skarnet joined
13:12 <fabled> hum
13:12 <fabled> doing ghc cross-compilation is slightly trickier than anticipated
13:13 <fabled> seems they by default just cross-compile a native compiler in one go
13:13 <fabled> the stage1 cross-compiler can be done separately
13:13 <fabled> but it's slightly tricky
13:13 <fabled> it expects the environment be cross-compiler for target; instead of for the host
13:14 <fabled> i wonder if i should just not support creating cross-compiler and do the full cross-build as native compiler
13:19 <ScrumpyJack> ncopa: apk search --exact -r so:libvncclient.so.0 shows remmina for me
13:19 <ScrumpyJack> apk 2.6.8 on armhf
13:23 <ScrumpyJack> also works on x86 but show nothing on x86_64
13:27 <ncopa> ScrumpyJack: because i pushed a rebuild for remmina
13:28 <ncopa> problem in repo is solved
13:30 BitL0G1c joined
14:51 <fabled> mitchty, where does bootstrap.patch come from?
14:53 hrmlgon-z joined
15:05 <fabled> mitchty, i'm at http://sprunge.us/cgQa but it still does not work because the cross-compiled native compiler is also prefixed with the target triplet (and paxmarking fails); need to figure out what to do there
15:07 <fabled> i also note that the previously when no --target was passed the ghc we now have things we are on x86_64-unknown-linux
15:07 <fabled> instead of -alpine-
15:09 <fabled> need to run, will continue later. please email me if you have ideas for removing the triplet from cross-compiled ghc
15:09 <fabled> i note that there was hackery to do it in the docker files of ghc-bootstrap
15:09 <fabled> bye
15:12 fabled joined
15:14 hrmlgon-z joined
15:30 fabled joined
15:45 Emperor_Earth joined
15:47 LouisA joined
16:01 <scadu> jirutka: pr just sent.
16:03 <fcolista> scadu, stupid question...but 3.6 is retro-compatible with 3.5?
16:08 <scadu> ffu, I see there are two conflicts. forgot to resolve them :s
16:10 <mitchty> fabled: i wrote it
16:10 <mitchty> you're hitting the same issue the rebootstrap project hits
16:10 <mitchty> and i hacked around in the dockerfile
16:11 <scadu> fcolista: afaik yes. bigger changes will be introduced in 3.7. sorry, I have shitty mobile connection right now which constantly disconnects.
16:11 <* fcolista> is scared that py3 packages stops to work after upgrade to 3.6 due to possible incompatibility
16:12 <mitchty> it should be in the ghc-bootstrap package
16:12 <mitchty> https://github.com/alpinelinux/aports/blob/master/testing/ghc-bootstrap/bootstrap.patch
16:12 <mitchty> we'd need to split it apart though
16:13 <scadu> fcolista: we should be fine :p
16:13 <mitchty> it does things that aren't quite right in the name of getting a stage 2 compiler that can build ghc
16:13 <fcolista> scadu, ok
16:13 <mitchty> the way the debian folks are doing things is they generate the stage 1 compiler, then ship that over to the target platforms for compiling on the target
16:13 <fcolista> i was looking to changelog and I don't see changes in the code syntax
16:14 <mitchty> ref https://groups.google.com/forum/#!topic/linux.debian.bugs.dist/q8f-Q59EQq0
16:15 <mitchty> and the hacks i did to get a native musl libc ghc to work were these https://github.com/alpinelinux/aports/blob/master/testing/ghc-bootstrap/Dockerfile.x86_64#L132
16:15 <mitchty> post build rather
16:15 <mitchty> i was going to tackle the makefile rules in the future to figure out where the triples were coming from
16:16 vakartel joined
16:16 <mitchty> ah he's gone, i should have kept reading
16:19 <* scadu> &
16:31 tty` joined
16:52 leo-unglaub joined
17:22 <kaniini> fabled: armhf triplet was wrong
17:23 mikeee_ joined
18:16 <fabled> mitchty, right. stage 1 basically the cross-compiler. i was thinking whether or not to package that too.
18:16 <fabled> mitchty, but can the stage2 by built without the triplet prefix if doing it in separate step?
18:17 <mitchty> fabled: i'm not 100% sure, every time i've tried it it always throws the triple in, and having gone through those makefiles its not been a high priority to fix but it should be doable
18:18 <fabled> do you think it's easier to fixup the $DESTDIR tree in package() or try to patch the makefile?
18:18 <fabled> imho, ghc got the cross-compiler stuff wrong
18:18 <mitchty> for now just linking stuff works like i did in the dockerfile
18:18 <fabled> well, it's just different from what gcc does
18:18 <mitchty> to be honest cross compilation isn't an oft used part of ghc
18:19 <mitchty> i'm being honest when i am saying I got told i was the first to use stage 2 binaries the way I did
18:19 <mitchty> it needs work to be sure
18:20 <mitchty> and for the bootstrap patch, we probably want to split that apart, i rip out a few things to make the bootstrap from glibc easier
18:20 <fabled> the problem is they tried to simplify the cross-compilation to single step, and broke the other use cases
18:20 <mitchty> aka the terminfo stuff
18:20 <fabled> i took the one hunk fixing the bootstrap
18:20 <fabled> i think terminfo works now that i made the makedepends_host correct
18:21 <mitchty> it should, note that only really affects colored terminal output from say the built in cabal stuff
18:22 <mitchty> but its normally there in ghc installs so likely good to keep it in
18:23 <mitchty> i can have a look at the makefiles probably tonight or tomorrow, or see if the debian guys have that sorted out yet
18:23 <mitchty> they hit the same deal with the rebootstrap stuff
18:24 <fabled> ok
18:24 <fabled> i'll probably continue on monday
18:24 <fabled> would love to get ghc done with
18:24 <mitchty> also as a note, i disagree with that madvise patch, i have a better way to fix the EINVAL https://gist.github.com/mitchty/916952e00f0d6f5cc7252ffe136bd27c
18:25 <mitchty> i've been running with it on some older kernels and an rts compiled on MADVISE_FREE kernel/libc, its much less annoying
18:25 <fabled> i would just try MADV_FREE if it fails with EINVAL; they fallback to using MADV_DONTNEED
18:26 <mitchty> i'll keep poking at it and work on upstreaming it
18:26 <fabled> permanently fallback that is
18:26 <fabled> the patch i took was committed upstream already
18:26 <fabled> but yeah, i agree it's not nice
18:26 <mitchty> yeah i've talked with them about it
18:26 <fabled> you get 2 syscalls there
18:28 <mitchty> yep, its not a huge problem as its not a hot path, but its an extra 100 or so instructions between user/kernel
18:30 <mitchty> also working on splitting out the profiled libraries into ghc-prof like in debian
18:30 <mitchty> that should help out most peoples install size and let the debugging stuff get installed if/when needed (most don't)
18:30 <fabled> right
18:32 <fabled> so the changes so far are simple as http://sprunge.us/URFD
18:33 <fabled> but needs the normalization of cross-built native compiler binary/path names
18:44 <mitchty> yep looks pretty straightforward
18:50 <mitchty> question for a split function as a sub package, it always seems to run before the primary package
18:52 t0mmy joined
19:05 <fabled> mitchty, package() should be run first, and the all the split function in the order listed in $subpackages
19:12 s33se joined
19:12 <mitchty> ok i'm doing something wrong then, i'll figure it out, i'm probably doing it wrong
19:17 fekepp joined
19:30 blueness joined
19:31 fekepp joined
19:36 mikeee_ joined
20:18 s33se_ joined
20:25 mikeee_ joined
20:28 s33se joined
20:40 s33se_ joined
20:59 <mitchty> and it was me doing it wrong, as always
21:08 <skarnet> the feeling of pebkac never gets old
21:12 fekepp joined
21:12 triple joined
21:16 <mitchty> fabled: also as a note, there is work on a new build system for ghc which will replace the makefile stuff, which means I have to probably help out with getting stage 2 builds working since I'm one of the few users
21:16 <mitchty> https://github.com/snowleopard/hadrian
21:39 triple joined
22:03 triple joined
22:17 triple joined
22:35 Emperor_Earth joined